TAAABLE 3 is a textual case-based cooking system which is a contestant in the third computer cooking contest, inheriting most features of its previous versions (TAAABLE AND WIKITAAABLE) and adding new features. The reused features are information retrieval and text mining for building a domain ontology and for annotating the ingredient part of the recipes, a semantic wiki in which the whole knowledge base (recipes, domain knowledge, retrieval knowledge, and adaptation knowledge) is encoded, and an inference engine based on minimal generalisations of the query and adaptation by ingredient substitutions. The two new features are related to adaptation. First, ingredient quantities adaptation is handled using conservative adaptation techniques on numerical constraints, permitting to maximally preserve, for instance, the quantity of sugar in the recipe. Second, adaptation is learned from a set of recipes using text mining techniques. Each recipe is analysed and represented as a tree which identifies ingredients, food components and actions that are performed. Formal concept analysis is used on these data in order to find an appropriate way to insert a new ingredient in the final recipe.